At HCA Healthcare, I led the architecture and production deployment of a multi-agent hybrid RAG platform for clinical workflows. It integrated EHR context, enterprise knowledge graphs, and vector retrieval, reducing clinician documentation time by 30–40% while maintaining HIPAA-compliant auditability.
I also designed A2A orchestration and an MCP layer for secure agent handoffs, and built GCP-native inference pipelines with intelligent on-premises and cloud routing. My work included retrieval improvements and evaluation practices to support safer, more relevant clinical assistant responses.
At Andor Health, I architected a multi-agent LLM framework for text and voice AI and introduced Pinecone retrieval with GraphRAG and Neo4j. The combined approach increased answer accuracy by 40%; a PEFT pipeline also cut fine-tuning time by 40% and reduced hallucinations in clinical workflows by 60%.
Earlier, at Hatch AI and IBM, I developed financial NLP and forecasting systems, document intelligence pipelines, recommendation engines, and healthcare classification models. Across these roles, I’ve worked from classical machine learning through production LLM systems in healthcare and financial services.
